New Research by Hodhod Examines Behavior of Socially Intelligent Agents

New research by TSYS School computer scientist Rania Hodhod and her colleagues from Princess Nourah bint Abdulrahman University, Ain Shams University, Cairo University and King Abdulaziz University extends prior work investigating effective ways to interact with socially intelligent agents in unpredictable environments.  Hodhod and her co-authors point out that behavior trees allow for knowledge to be represented in graphical formats that provide a way for socially intelligent agents, which are agent systems that are able to connect and interface to humans (e.g., robotic systems, computational systems), to effectively interact with received information.  The studies explains that behavior trees are capable of storing prior social experiences that can be used by socially intelligent agents to provide adequate human-like interactions when facing new social situations.  According to Hodhod, “one challenge appears when a social agent with vast past experiences – represented as a forest of behavior trees – tries to retrieve a similar behavior tree to learn from in order to provide plausible interactions in the current situation in a cost-effective manner.  With their inherent temporal structure, cognitive scripts can facilitate the use of contextual retrieval techniques on behavior trees.”  This study, which appears in the latest issue of Electronics, introduces novel hybrid retrieval techniques that group behavior trees, represented as cognitive scripts, into compact clusters that can then be used to retrieve the most similar behavior tree to a query in real time without noticeable delay.  Examination of the performance of the proposed hybrid-retrieval technique using a semi-structured dataset of cognitive scripts indicates that the technique proposed by Hodhod et al. is more cost-effective than other clustering techniques as it produced a shorter average retrieval time.